Treasury forecasts: the tendencies and consequences of Inaccuracy


In a briefing paper released today, I examine the tendencies and accuracy of treasury forecasts for GDP growth contained in the spring budget. My sample contains three different types of forecast: Type I are forecasts for the same year as the budget, Type II are forecasts for the next year and Type III are forecasts for the year after next. All are made simultaneously in each spring budget. Using some basic statistical and probabilistic analysis, I reach three key conclusions based on my sample of thirty forecasts:

  • There is a low correlation between forecast and outturn for Type II and III forecasts.
  • All three types of forecasts are less likely to be accurate when economic growth is changing faster in either a positive or negative direction.
  • There is strong evidence of a bias towards overestimation in each type of forecast

Statistical analysis can sometimes be quite dry, so I think it is important to spell out the significance of the findings. First, and most obviously, it seems we should place little trust in the predictive power of the Type II and III forecasts given their low correlation with outturns. Second, and perhaps more significantly, the combination of a forecasting process that does very badly at foreseeing higher growth in either a positive or negative direction, and the finding of strong evidence for a bias towards overestimation, means that treasury fiscal projections for future spending and borrowing plans (both elastic to changes in growth) can be highly inaccurate and unrealistically optimistic.

For example, in 2007 the treasury was predicting 2009 growth at positive 2.5%. In fact, the economy shrank by 5% in 2009. The projections for the national debt and deficit, which are given as a percentage of GDP, were accordingly way off the mark. This seriously undermines any attempt at fiscal discipline.

I argue that to counteract the effect of such inaccurate projections, government should use very pessimistic assumptions about future growth when projecting debt and borrowing. Currently, the treasury seems to place quite a lot of confidence in its forecasts, adjusting the figure down by just a quarter per cent when it plans future borrowing. The data suggests that this figure should be far larger. I don’t go as far as to suggest an actual figure: perhaps that is for a different paper.